itojun said:
 > 	if my understanding is correct,
 > 	- some of Hitachi ICE and other debugging tools are using symbols
 > 	  without underscores

???

Basically, old gcc for sh-elf uses symbold *WITH* underscore.
It's differnt from the normal ELF spec.

And recently, gcc-current switches NOT to use underscore.
egcs-2.95.2 still uses underscore.

itojun said:
 > 	- gcc/binutils is synchronized to the above
 > 	so if we add underscores, we cannot use those debugging tools.
 > 	msaitoh should have more details...

If we switche not to use underscore, some ICEes may confuse (not verified
because I'm using COFF only).

Todd said:
 > A prior response from (msaitoh?) indicated that this convention was chosen
 > so that assembly code would conveniently work between ELF and COFF
 > executables unchanged.  A lazy thing on the part of the eval-board/toolchain

I didn't say such a thing :-)

 Some informations:

	1) In September 2000, sh-elf specific part of gcc-current have
	  not to use underscore and to use GOT.

	2) At least, Linux/SH switches not to use underscore (uses above).

	3) egcs-1.1.2 doesn't supports -m4 option (i.e. SH4)
 
	4) egcs-2.95.2 supports -m4 (this version doesn't contains 1))

	5) sh's egcs-2.95.2 is tooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo buggy. Yes offcourse gcc-current is buggy, too.

So, I think 4)+1) is better.

----------------------------------------------------------
		SAITOH Masanobu (masanobu@iij.ad.jp
				  msaitoh@netbsd.org)